4  Communication

library(tidyverse)
library(ggpp)
library(ggrepel)

4.1 Plotting

4.1.1 ggpp

The ggpp package allows natural placement of text in ggplots based on position within frame instead of variable values. This is an extension to the Grammar of Graphics (r-project.org).

The key function here is geom_text_npc, as in this example:

df.lab = tibble(x = c("left", "right"), y = c("top", "bottom"), 
                label = c("High power,\nlow displacement", "Low power, high displacement"))
gp1 = ggplot(mtcars, aes (x = disp, y = hp, col = as.factor(cyl)))+
         geom_point()+
  theme_bw()+
  geom_text_npc(data = df.lab, aes(npcx = x, npcy = y, label = label))
gp1

The big advantage here is that the text is placed based on the window rather than x or y coordinates, so changinge the data or x or Y limits doesn’t screw up text placement in relation to the window.

gp1 + ylim(c(-100, NA))

When adding text directly (not from dataframe), ggpp overwrites the annotate function to support natural placement.

gp1 + 
  ylim(c(-100, NA))+
  annotate(geom = "text_npc",
           npcx = "left",
           npcy = "bottom",
           label = "New text")

ggpp is also great to inset tables (geom_table()), plots (geom_plot()) or images (geom_grob()) within each plot.

One way to inset a table with geom_table(), which takes a tibble in which one object is itself a list of the desired table contents. This feels a little fiddly to me, but probably scales up well when making many figures.

df = mtcars |> 
  group_by(cyl) |> 
  summarize(n = n())
df.inset = tibble(x = 400, y = 300, tb = list(df))
ggplot(mtcars, aes (x = disp, y = hp, col = as.factor(cyl)))+
         geom_point()+
  geom_table(data = df.inset, aes(label = tb, x = x, y = y))+
  theme_bw()

We can also control this more “naturally” with the overwritten annotate function – no need to make a tibble containing a tibble or specify locations in absolute x and y coordinates.

ggplot(mtcars, aes (x = disp, y = hp, col = as.factor(cyl)))+
         geom_point()+
  annotate("table_npc", npcx = "right", npcy = "top", label = df)+
  theme_bw()

This same annotate approach can be used to add a plot inset into the figure. Make the inset plot and assign it to a variable, then use annotate with “plot_npc” like so:

gp.inset = ggplot(mtcars, aes(x = mpg, y = hp))+
  geom_point()+geom_smooth()+
  theme_bw()
ggplot(mtcars, aes (x = disp, y = hp, col = as.factor(cyl)))+
         geom_point()+
  annotate("table_npc", npcx = "right", npcy = "top", label = df)+
  annotate("plot_npc", npcx = "left", npcy = "top", label = gp.inset)+
  theme_bw()

`geom_smooth()` using method = 'loess' and formula = 'y ~ x'

4.1.2 ggrepel

The ggrepel package can be used to annotate data points using repultsion to minimize point overlap.

The problem:

dat = mtcars |> 
  head(10)
dat$name = rownames(dat)
ggplot(dat, aes(x = disp, y = hp, col = as.factor(cyl)))+
  geom_point()+
  geom_text(aes(label = name), col = 'black')

The solution:

ggplot(dat, aes(x = disp, y = hp, col = as.factor(cyl)))+
  geom_point()+
  geom_text_repel(aes(label = name), col = 'black')

4.1.3 Custom labeling

Sometimes we want to add content to ggplot in different layers, but have a legend that correctly represents these layers. The solution is to add the desired graphical setting (e.g., color, shape, etc) in the aes() call of each layer, and assign a character atomic to that argument. This will then distinguish the graphical setting just like when we assign a character vector column of the dataframe to the graphical argumnet. The lab() function can be used to give a meaningful legend title, and the standard modification functions (e.g. scale_color_manual) can be used to specify the colors (or shape, or size, etc) associated with each layer.

ggplot(mpg, aes(displ, hwy)) + 
  geom_point() + 
  geom_smooth(aes(color = "loess"), method = "loess", se = FALSE)+
  geom_smooth(aes(color = "gam"), method = "gam", se = FALSE)+
  labs(color = "method")

`geom_smooth()` using formula = 'y ~ x'
`geom_smooth()` using formula = 'y ~ s(x, bs = "cs")'

4.2 Interactives

4.3 Tables

4.4 Flowcharts

Quarto has built-in support for generating flow charts and other diagrams from markdown. See here: Quarto - Diagrams. This uses mermaid and/or GraphViz. For live building of these tools, see:

• Online FlowChart & Diagrams Editor - Mermaid Live Editor
• Graphviz Online (dreampuf.github.io)

Here’s the default chart of the mermaid live editor as an example

flowchart TD
    A[Christmas] -->|Get money| B(Go shopping)
    B --> C{Let me think}
    C -->|One| D[Laptop]
    C -->|Two| E[iPhone]
    C -->|Three| F[fa:fa-car Car]